Also, I find it weird to put agricultural and industrial revolution together in this context.
Agricultural revolution was when people shifted from doing less physical work (hunting) to doing more physical work (building a civilization). It probably wasn't even by choice, but simply because agricultural societies were more efficient and would conquer those which didn't settle down. About the only thing I see here in common with industrial revolution is the forced change of lifestyle.
Perhaps you mean the initial spark of an invention, the idea which strikes like a lightning, and so on? But many inventions happen by accident (like the post-it note invention, http://www.todayifoundout.com/index.php/2011/11/post-it-note...). And then you could argument that only human are crazy enough to find novel usage for useless things.
I don't know.
Or is it a gradient, where the high end of the mental work will remain performed by humans, and will grow to encompass more work?
And if it works like that, can our education systems change to keep people productive for longer?
Also, it may be one with a solution that's easy to state - but not necessarily easy to reach. We can't suddenly change how the economy works - we don't have direct control over it. We can try to transition the current one to the desired future one my means of setting up incentive structures appropriately.
I don't know what would or could prevent that assuming the ratio and rough magnitude between those two figures. At that point, you'd hope that the pursuit of status would presumably move elsewhere.
Engineering incentives is beyond difficult. Many, many of the things people complain about most in the modern economy is due to some crufty legacy incentive-engineering left over from years past.
Once most of customer service and human drivers are eaten by software, which will probably happen without solving 'hard AI' tech will employ even fewer people.
Jobs will only be created out of that if cheaper transportation and customer service enables a new place in the market that wasn't already existent.
So far I don't know of any jobs that the AI revolution has created other than people that create and train AI systems. It simply wouldn't be profitable to employ these people if they weren't destroying more jobs than they take up.
A hundred farm workers get replaced by one tractor, but that's okay, because those hundred workers become tractor repairmen, tractor factory workers, tractor designers, and we can farm far more land and produce wealth more efficiently. But not everyone who could plow a field can design a tractor. A thousand years ago, damn near every job was manual labour and could be done with a two-digit IQ, but we're reaching the point in first-world countries where less than half the work which needs to be done can actually be done by someone in the lower half of the population's intelligence distribution.
If we're not there yet, we will be soon. And I think it's fundamentally different to every previous automation revolution.
If we pushed towards educating people in practical skills, like tractor design, do you really think that it would be a large number of people who are incapable of grasping the concepts? Especially if we used the modern research we have into how people learn effectively?
Education as an industry and cultural norm needs to grow -- and not in the abominable way that it is now, with us glorifying education for the sake of education -- but a push towards valuing people who have learned specialized skills, regardless of what they are.
There's a reason that we have a dearth of mechanics in the U.S., and it's not because being a mechanic is an unattainable goal, but rather because the social norms around specialized blue collar work are ridiculous in the U.S.
Besides we still haven't really figured out post-industrial revolution employment all that well. Lots of people in bullshit jobs collecting paychecks just because.
Before the industrial revolution we had lots of horses for horse power. Then we made engines which replaced horses. Horses didn't change jobs, they couldn't.
Low hanging fruit here is education, tradespeople (AI will not be repairing your car for a while), managerial and oversight tasks, QA engineers, blah, blah, blah. There are plenty of things that humans will be doing better than AI, and from those we'll probably continue specializing and be more ready to handle when automation of more complex tasks does eventually happen.
That may not be the right question. There are all kinds of things that theoretically we could replace with bots over humans but we do not because of convention. For example Home Depot or any Home/Shed tool supply store. Let me explain?
In Neal Stephenson's The Diamond Age there is a class of humans who generate handcrafted items. They have machines that produce anything to molecular perfection and yet humans (due to our nature) seek to distinguish themselves from other groups of humans using objects. In this case flawed objects, but objects they willingly paid for nonetheless. It's all about status!
This is not outlandish to me. I want to construct a Tiny House. I know that when I develop the skills and accomplish this task I will place considerable value on my creation.
It is extremely likely that this is not optimal because of the division of labour. However I will enjoy my Tiny House regardless.
AI can endlessly optimize production but it cannot manufacture personal pride or social class.
Human nature creates scarcity naturally even within a state of abundance.
We really do like to belong to tribes and be 'self made'.